In his book 'How progress ends', Dr. Carl Benedikt Frey warns that innovation and not automation is more likely to hold the key to future success.

“The hope right now for many is that AI is going to revive productivity growth going forward, and you might argue that the kind of consolidation that we've seen in many industries would be good for growth in AI, if scaling was all you needed. But that's not what we've been seeing,” Frey says.

More resilient models are needed

At the event ‘Saviour or stagnator? Technology, AI and economic growth’, hosted by the Resolution Foundation, Frey pushed back on the narrative of the correlation between technological advancement and growing economic prosperity. Frey’s new book looks back over the past millennium and uses these lessons from history to look ahead to the impact of AI on growth.

“Between 2019 and 2024 we’ve seen large language models scale by 10,000x and we saw very little improvement in performance on basic reasoning tests. What we need is models that are more resilient to changing circumstances, and that means that we need more innovation.”

“Business dynamism is not just a drag on innovation, it also shapes the direction of innovation,” Frey adds. “Large firms have scale, so they're more likely to invest in process, innovation and automation. Smaller firms don't have scale, so they're much more likely to invest in products that create new types of industries and new types of work. And I think that's the key driver of job creation, growth and prosperity.”

Lowering barriers to entry

Frey cites the example of professional services where use of generative AI is, broadly speaking, reducing barriers to entry. “It's a bit little bit like what Uber did to taxi services with GPS technology; knowing the name of every street in London was no longer a particularly valuable skill, and with the app matching supply and demand, anybody with a driver's licence could get into his or her car and top up their incomes.”

The difference with professional services is that those are traded, “so if AI reduces the productivity differential between a worker in Manila and a worker in London, and the wage differential between the two is an order of magnitude, I'm more optimistic about those places for catch up growth,” Frey says.

“Generative AI is not quite a fully-fledged automation technology, because you need a human in the loop for most activities. So, I think of it more as reducing barriers to entry in relatively skilled work. I think there's going to be a lot of changes because of AI. I particularly worry about the loss of status of people working in professional services.”

Reflecting on The Future of Employment

In 2013, Frey co-authored a working paper 'The Future of Employment', grabbing headlines with its prediction that 47% of US jobs would be lost to automation. Reflecting on the paper, Frey is less sensationalist in his outlook.

“I don't think we're about to have mass unemployment. I'm totally open to that AI could deliver something like the productivity upsurge we saw during the ICT revolution, but that was not particularly long sustained, and actually quite disappointing compared to what we saw in the post-war years in terms of sustaining productivity growth.”

Frey also points to data that shows firms with more than 250 employees scaling back AI use, which he says gives him some cause for concern.

“There are a lot of bottlenecks in various industries that are hard to overcome, and most of the use cases I see are about automation and process improvement. I wouldn't exclude the possibility that we are at the cusp of something new. But I don't quite see the foundations for making that sustainable.”